← back
Designing AI To Scale Human Thought — Jun Yu Tan, Tusk
Original: Designing AI To Scale Human Thought — Jun Yu Tan, Tusk
Takeaway
AI products that augment human thinking (blind-spot detection, cognitive partnership, proactive guidance) beat those that automate the human out of the loop.
Summary
- Tusk argues for augmentation over automation: AI should help users produce high-quality work, not replace them — especially in high-judgment domains like coding.
- Three core interaction patterns: blind-spot detection (systematic pessimism), cognitive partnership (theory-of-mind about how users think), proactive guidance (timing matters most).
- Tusk's product generates and executes unit tests against PRs; surfaces verified bugs caught in 43% of PRs at enterprise customers (DeepLearning.AI, Teampay), added ~1000 new tests in 2 months.
- Novelty/criticality framework: prioritize high-novelty+high-criticality (race conditions, exposed data) as interruptions; batch low/low into optional polish suggestions.
- Trust must be progressive, contextual, and bidirectional — like a new team member, not an offshore contractor.
augmentationproduct-designcode-review
Original description
Forget the hype of AI automation replacing jobs. The future lies in human augmentation — revealing blind spots, sparking creativity, and amplifying thoughtful decision-making. In this talk, we’ll explore the principles that distinguish augmentation from automation in AI UX design, covering interaction patterns, design principles, and trust-building feedback loops. Drawing from real-world experiences building AI-powered tools and beyond, we’ll dive into concepts for crafting interfaces that empower users to think smarter, not just work faster. Expect practical insights and a fresh perspective on AI’s role as a collaborative partner. AI Augmentation: https://jytan.net/blog/2025/ai-augmentation/ Tusk: https://www.usetusk.ai/